- Ground-motion model is: where Y is in g, Mh = 7.0; a1 = 0.44780, a2 = 0.24582, a3 = -0.14444, a4 = 0.49645, b1 = -1.17792,
b2 = 0.04959, b3 = 0.00000, h = 4.52478, c1 = 0.68185, c2 = -0.10727, Δb3,Alborz = 0.00000, Δb3,Zagros =
0.00000, Δ3,Others = 0.00000, τ = 0.20592 (inter-event), ϕS2S = 0.20338 (site-to-site), ϕ0 = 0.45542
and σ = 0.53961 (total) for horizontal PGA; and a1 = -0.32176, a2 = 0.00795, a3 = -0.14011,
a4 = 0.32612, b1 = -1.60377, b2 = 0.12555, b3 = -0.00223, h = 4.90710, c1 = -0.01229, c2 = 0.00132,
Δb3,Alborz = 0.00000, Δb3,Zagros = 0.00000, Δ3,Others = 0.00000, τ = 0.21104 (inter-event), ϕS2S =
0.16597 (site-to-site), ϕ0 = 0.50317 and σ = 0.57032 (total) for vertical PGA.
- Characterise sites using V s,30. Data from 321 different stations. Data has 155 ≤ V s,30 ≤ 1564m∕s but only
recommend use for 300 ≤ V s,30 ≤ 1000m∕s.
- Use data from Iranian Strong Motion Network from 1979 to 2013. Data from SMA-1 and SSA-2
instruments. Visually inspect data and remove poor-quality records. Baseline correct records. Remove
instrument response. Individually bandpass filter zero-padded records using phaseless 8-pole filter using
cut-off frequencies determined for when signal-to-noise ratio is < 3.
- Exclude data that:
- Lack all 3 components;
- Have a focal depth ≥ 35km;
- Have a rjb ≥ 250km;
- Do not have known V s,30 (30% of original data have no V s,30);
- 46 earthquakes have only a single record. Note that this may mean τ and ϕS2S are slightly underestimated.
- Note that few data from rjb < 10km so model not well constrained for short distances.
- Include terms to account for potential differences in anelastic attenuation between: Alborz (132 records),
Zagros (262 records) and other (e.g. central and eastern Iran) regions (294 records).
- Use nonparametric plots of data against, e.g., rjb, Akaike and Bayesian information criteria and try various
trial functional forms to choose final functional form.
- Allow for oversaturation of ground motions for large magnitudes within functional form.
- Constrain b3 to be zero or negative to avoid unrealistic distance scaling and because of trade-off between
geometric and anelastic terms.
- Do not account for nonlinear site effects because dataset is deficient in records with high Mw, short rjb
and low V s,30.
- Originally try to model regionality in c1 and c2 but effect is not significant.
- Do not include fault mechanism in model because total σ reduced by 10% by removing this term and effect
of inclusion of these terms is statistically insignificant.
- Regress to obtain source and path coefficients and then regress again to find site coefficients.
- Examine residuals (unbinned and binned: using 0.25Mw, 100m∕s and 25km intervals, computing mean
and 95% confidence intervals when ≥ 3 records). Find no trends.